Modern computing often requires the collection, processing, or storage of very large data sets or file systems. Accordingly, to accommodate the capacity requirements as well as other requirements, such as, high availability, redundancy, latency/access considerations, or the like, modern file systems may be very large or distributed across multiple hosts, networks, or data centers, and so on. Conventionally, some distributed file systems implement various pre-fetching schemes to help improve read performance. However, implementing a pre-fetch policy that consistently or predictably improves read performance may be difficult because of the variability of client read patterns. Also, a per-request pre-fetch policy scoring that enables a file system to determine if pre-fetching should be executed may be challenging to implement at scale because of vast number of file system requests that must be handled. Thus, it is with respect to these considerations and others that the present invention has been made.